The Ting’s rice core collection, i.e., a total of 150 accessions of rice landraces, was used to screen their salt tolerance (Supplementary Table S9). These landraces were mainly collected from 20 different provinces in China as well as from North Korea, Japan, Philippines, Brazil, Celebes, Java, Oceania, and Vietnam. These regions are distributed across the north latitude 55° to south latitude 10° and including regions with temperate, tropical, and subtropical climate. The core collection was constructed from 150 accessions of 2,262 based on a strategy of stepwise clustering and preferred sampling on adjusted Euclidean distances and weighted pair-group average method using integrated qualitative and quantitative traits (Li et al., 2007). Of the 150 landraces, 32 were classified as japonica rice (24 were typical japonica rice, and 8 were japonica-clined rice), and 118 were classified as indica rice (16 were indica-clined rice, and 102 were typical indica rice), according to Cheng’s index criterion (Li et al., 2011).

The seeds of the 150 landraces were placed in an oven at 50°C for 48 h to break dormancy. From each landrace, 40 seeds with uniform size and full rice grains were selected, and then the seeds were soaked in 75% alcohol for 15–20 min for disinfection treatment. Then, the seeds were rinsed with sterile water for three times. The rinsed seeds were put in a net bag, soaked in distilled water, and placed in a thermostat at 30°C for 48 h to incubate germination. Then, the seeds were transferred to 96-well PCR plates with cut wells and distilled water. The PCR plates were placed in a culture room with light at 28°C for cultivation and later cultivated with 12 h of light and 12 h of darkness. After 7 days, the distilled water was changed to the standard nutrition of the International Rice Research Institute (IRRI) but with only 0.5-fold concentration (Yoshida et al., 1976), while keeping the PH value of Yoshida’s solution at 5.5. The nutrient solution was changed once per 3 days. After culturing for 7 days, the cultivation was changed to 0.5-fold Yoshida’s solution and later changed to one-fold Yoshida’s solution with regularly replacing the nutrient solution. When the seedlings grew to the three-leaf stage, the samples of both the control and the salt treated were extracted in 25 ml acetic acid (100 mm) at 90°C for 2 h, and 2 ml extraction was divided into two groups for sodium and potassium, respectively. 100 mm NaCl solution was added to the nutrient solution for salt stress treatment, while the control group continued to grow in the nutrient solution. Each treatment was set with three repetitions. After 10 days of treatment, samples were taken to measure the phenotypic traits. Shoot and root Na⁺ and K⁺ concentrations (RNC, SKC, and SNC) of each sample were determined by atomic absorption spectrometry (AAS, Series 2, Thermo Electron Corporation). Concentrations of sodium and potassium in shoots and roots were expressed in millimoles per gram (mMg⁻¹; Qi et al., 2005; An et al., 2020). The K⁺/Na⁺ ratios in roots (RN/K) were calculated subsequently. A root scanner (Expression 1100XL) was used to analyze root traits. To reduce errors, 10 seedlings were scanned for each material, and each seedling was scanned 3 times, and the average values of TRSA (total root surface area), TRV (total root volume), and TRL (total root length) were calculated. Set up 3 biological replicates. RTRSA (relative total root surface area), RTRV (relative total root volume), RTRL (relative total root length), and RSN/K (relative ratio of Na⁺ to K⁺ concentrations in shoots) were calculated according to the following formula: relative trait value (%) = (trait value under salt stress) / (trait value under control) × 100.

In this study, to further determine whether candidate genes are related to ST, we first screened extreme salt-tolerant (S125) and extreme salt-sensitive (S87) rice landraces, RT-PCR of candidate genes was performed in salt-tolerant (S125) and sensitive (S87) rice landraces. Under the condition of salt stress, the time for rice seedlings to appear stress phenotype lags behind the response time of related genes. Under salt stress, the expression levels of some salt-related genes will change rapidly and then return to their original levels, resulting in undetectable changes in expression levels. Using high concentration and short time salt stress for gene expression analysis can more accurately analyze the changes of related gene expression. Therefore, in this study, the 20-day-old seedlings were treated with 200 mm NaCl for a short time (0, 3, 6 and 12 h) for gene expression analysis (Xiaoxue et al., 2019). The expression of genes at 0 h of salt stress in this study is as control data for non-stress (without salt stress). The total RNA from the shoots tissues of landraces was extracted using steadyPure Plsant RNA Extraction Kit (Accurate Biotechnology; An et al., 2020). The cDNA for real-time PCR was reverse-transcribed from 2 μg of total RNA using Evo M-MLV Reverse Transcription Reagent Master Mix (Accurate Biotechnology). According to the manufacturer’s instructions, real-time PCR was performed using SYBR Green Pro Taq HS kit (Accurate Biotechnology) in a real-time PCR machine (QuanStudio). Relative gene expression levels were determined using the 2^-ΔΔCt method (Livak and Schmittgen, 2002). Data analysis and graphing were performed using GraphPad Prism 6.02 software, and Duncan’s test was performed using SPSS. The rice actin was selected as the internal control. Primers used for qRT-PCR analysis are listed in (Supplementary Table S10).

DNA Quick Plant System (Tiangen Biotechnology, Beijing, China) was used to extract plant genomic DNA. The target sequence was amplified with specific primers. The amplified product was purified with EasyPure PCR purification kit (Tiangen Biotech, China) and quantified with NanoDrop 8,000 spectrophotometer (Thermo Fisher Science, Waltham, MA, United States). Sequence alignment was performed with the DNAMAN software using the genes in the Nipponbare genome as a reference.

GWAS was performed by using the compressed MLM program in the Tassel5.0 software, where the model was as follows: Y = Xα + Qβ + Kμ + e, where Q represents the kinship, X is genotype, Y is phenotype, while Xα and Qβ considered as fixed effects and Kμ and e as random effects. The population structure K was calculated by the Admixture software, and the kinship between samples was calculated by the SPAGeDi software. The significant SNPs were identified by p value. To obtain high-density SNPs, we encrypted the original 67,511 SNPs and obtained 2,487,353 SNPs by the following method: the user uploads input files in oxford format (.gen/.sample) per chromosome. The imputation server utilizes chromosome recombination maps and the Rice Reference Panel (RICE-RP) haplotypes to impute the user’s data out to 5.2 M SNPs with IMPUTE2 and returns imputed results in plink binary format. The user can filter the imputed data using plink1.9 to produce a final data set of desired SNP density and composition (Wang et al., 2018b).¹ Manhattan scatter plots and QQ plots are drawn using the “qqman” package in R software. LD heatmaps surrounding peaks in the GWAS were constructed using “LD heatmaps” in the R package (Shin et al., 2006). The haplotypes of at least three rice landraces were analyzed for phenotypic comparison. Differences in phenotypic values between alleles of each non-synonymous SNP were assessed by Student’s t tests. The sequence alignment of each gene was determined using non-synonymous SNPs associated with ST, and differences in phenotypic value between haplotypes of each gene were calculated by one-way ANOVA or Student’s t tests.

In order to evaluate the phenotypic variation of ST in the core collection at the seedling stage, a statistical analysis was performed on 8 ST-related traits: RNC (root Na⁺ concentration), SKC (shoot K⁺ concentration), SNC (shoot Na⁺ concentration), RN/K (ratio of Na⁺ to K⁺ concentrations in roots), RTRSA (relative total root surface area), RTRV (relative total root volume), RTRL (relative total root length), and RSN/K (relative ratio of Na⁺ to K⁺ concentrations in shoots; Supplementary Table S1). The results showed that all traits showed tremendous phenotypic variation in the population. In particular, RN/K had the highest coefficient of variation, and RTRSA had the lowest coefficient of variation. Correlation analysis showed that most traits were significantly positively correlated. Significant negative correlations were observed only between SKC and RSN/K, SKC, and RN/K (Supplementary Table S2). The results showed that salt stress has different degrees of influence on the 8 ST-related traits at rice seedling stage.

In previous studies, the rice core collection was sequenced using the Specific-Locus Amplified Fragment Sequencing(SLAF-seq)approach and 67,511 SNPs were obtained (Zhao et al., 2018). Based on them, 2,487,353 SNPs were further obtained by the imputation method of (Wang et al., 2018b). With consideration of the population structure and kinship, the MLM (+Q + K) model and 2,487,353 SNPs were used to perform GWAS on eight traits related to ST with the high density SNPs set, and the results are presented in the form of Manhattan plots (Figure 1) and QQ plots (Supplementary Figure S1). With a significant threshold of p < 0.0001, 843 significant SNPs were detected (Supplementary Table S3), which were unevenly distributed on the 12 chromosomes. The most significant position with the largest contribution rate is located on chromosome 1, which can explain 28.11% of the phenotypic variation.

ST-related QTLs are defined by the decay distance of linkage disequilibrium (LD). Previous studies had shown that the decay distance of LD is 500 kb (Zhao et al., 2018). Therefore, a region was considered as one QTL if it had more than one SNP with p < 0.0001 within the LD decay distance. We named salt tolerance QTLs with reference to the method proposed by Mccouch et al. (1997). In total, 65 QTLs were identified significantly associated with ST (Figure 2). There are 1–13 of QTLs on each chromosome, and each QTL contains 1–379 SNPs. These QTLs explained 13.47 to 28.11% of the phenotype variation (Supplementary Table S4).

For RNC, we detected 8 QTLs located on chromosomes 1, 3, 5, 8, 9, 10, and 12 under salt stress, which could explain 14.83–23.96% of the phenotypic variation. Under salt stress, we detected 14 QTLs for SNC on chromosomes 1, 2, 3, 5, 6, 7, 9, 11, and 12, which can explain 13.91–23.61% of the phenotypic variation. For RN/K under salt stress, ten QTLs were detected on chromosomes 1, 3, 5, 6, 8, 9, 10, and 11, which could explain 13.75–27.74% of the phenotypic variation. For RTRV under salt stress, 8 QTLs were detected on chromosomes 3, 6, 8, 9, and 11, which could explain 13.47–20.52% of the phenotypic variation. Under salt stress, we detected 13 QTLs for SN/K on chromosomes 1, 2, 3, 4, 5, 8, and 9, which could explain 13.67–21.14% of the phenotypic variation. For SKC under salt stress, two QTLs were detected on chromosomes 6 and 8, which could explain 14.38–17.94% of the phenotypic variation. Under salt stress, we detected 5 QTLs for SKC on chromosomes 1 and 12, which can explain 14.46–28.11% of the phenotypic variation. For TRSA under salt stress, 5 QTLs were detected on chromosomes 1, 7, 9, 10, and 11, which can explain 14.33–23.10% of the phenotypic variation.

Through association mapping analysis, there are 6 genomic regions containing co-localization QTLs. We defined co-localization QTLs as qTL1, qTL2, qTL3, qTL4, qTL5, and qTL6, respectively (Table 1). The most complicated is the co-localization region of chromosome 6, i.e., qTL4, which is composed of 3 QTLs and has a large overlap region. The remaining 5 co-localization regions are all composed of 2 QTLs.

Since qTL4 is composed of 3 QTLs with 14 significant SNPs, it was chosen for further analysis. The qTL4 contains the highest peak SNP (Chr6_28930159) at approximately 28.93Mbp, which can explain 14.38–17.94% of the phenotypic variation. According to the LD decay analysis of the population, the 250 kb upstream and downstream around the peak SNP were designated as the searching range of candidate genes as shown by the LD heat map (Figure 3). Through the Rice Annotation Project database,² ninety-four genes were found in this region (Supplementary Table S5). It includes 50 functional annotation genes, 24 expressed proteins, 17 transposon proteins, and 3 hypothetical proteins. Among them, a known salt tolerant gene (OsSIDP366) was found in this region, which was located in the interval between qSKC.6 and qSNC.6. Besides it, qRN/K1.1 and qRSN/K1.1 also were co-localization, which were detected in the interval of qTL6 and related to ratio of Na⁺ to K⁺ concentrations. The qTL6 contains the highest peak SNP (Chr9_19518843) at approximately 19.51Mbp, which can explain 13.75–16.92% of the phenotypic variation. In addition, we also found a SNP Chr6_28822519 in the qRN/K.6 interval with a significant threshold of p = 0.000069293, which could explain 17.49% of the phenotypic variation, indicating that there may be new salt resistance-related genes in this interval. Based on gene functional annotation and GO enrichment analysis³ (Supplementary Table S6), we screened and found 11 out of 50 functional genes which are related to stress response or metabolism process-related genes were salt tolerance-related candidate genes (Supplementary Table S7).

We further performed haplotype analysis on the non-synonymous mutant SNPs and the promoter SNPs in the exon region of the 11 candidate genes and detected significant differences among different major haplotypes based on the haplotypes with all non-synonymous SNPs. It was found that the haplotypes of the 4 candidate genes had significant differences (Table 2), and each candidate gene had a different number of non-synonymous SNPs (Supplementary Table S8). The haplotype analysis showed that the four candidate genes were significantly associated with SKC (shoot K⁺ concentration), and the SKC of different haplotypes were also significantly different (Figure 4). These findings suggest that the four candidate genes (LOC_Os06g47720, LOC_Os06g47820, LOC_Os06g47850, and LOC_Os06g47970) may be involved in the regulation of salt tolerance at seedling stage in rice.

To further determine whether the aforementioned 4 candidate genes are related to ST, gene expression assays for the 4 candidate genes were conducted in salt tolerant (S125) and sensitive (S87) rice landraces. We performed a statistical analysis on the dead seedling rates in S87 and S125 under salt stress to further verify their differences in salt tolerance (Figure 5). There were significant differences in the dead seedling rates of S87 and S125 under the two salt stress concentrations, indicating that they had significant differences in salt tolerance. Therefore, S87 and S125 were selected for subsequent analysis of candidate genes expression. The gene expression level of LOC_Os06g47720 in S87 showed a continuous increase trend from 0–12 h and was higher than S125 for all tested time points. The gene expression level of LOC_Os06g47720 in S87 was nearly 10-fold higher than that in S125 after 12 h of salt stress treatment. The opposite expression pattern was observed for LOC_Os06g47820, LOC_Os06g47850, and LOC_Os06g47970, where the gene expression levels in S125 were higher than those in S87 for all tested time points after salt stress treatment. Among them, the gene expression levels of LOC_Os06g47850 and LOC_Os06g47970 in S125 were nearly 10-fold higher than those in S87 after 12 h of salt stress treatment.

To determine the association between candidate genes and ST phenotypes, four candidate genes were sequenced for the two rice landraces S125 and S87 (Supplementary Figure S1) and the sequence alignment was generated using DNAMAN. LOC_Os06g47720 has 29 mutation sites in the two ST extreme genotypes, which lead to 11 amino acids changed in the S125, while translation was terminated early in the S125. For LOC_Os06g47820, it showed that the two genotypes had 17 mutation sites and 14 amino acids changed. LOC_Os06g47970 has a pair of base substitutions in S125, which caused the glycine to change into the asparagine. LOC_Os06g47850 has no sequence difference between the two genotypes.

Artificial directional domestication and selection have led to the reduction of genetic diversity in many crops such as rice, and many excellent resistance gene resources have been gradually lost in common cultivated rice. Some landraces are less affected by artificial selection and contain more genetic diversity and resistance genes. The core collection of rice landraces was constructed from a total of 7,128 accessions of rice landraces from the Ting’s rice collection (Li et al., 2011) and came from all over China as well as some main rice cultivation countries. It provides an important gene pool of genetic diversity and beneficial genes for rice breeding. For example, several QTLs for aluminum tolerance and cold tolerance were identified in the core collection (Song et al., 2018; Zhao et al., 2018). In this study，8 traits related to ST showed tremendous phenotypic variation in the core collection, which suggested that salt stress has different degrees of influence on the ST-related traits at the seedling stage, and these phenotypes could be used for GWAS. Furthermore, a total of 65 QTLs significantly associated with salt tolerance were identified by GWAS. These results further verified that the core collection of rice landraces contains abundant resistance gene resources to biotic stress and abiotic stress and is a precious germplasm resource with great prospect for mining and utilization of beneficial genes for rice breeding.

GWAS is a promising method for QTL fine-mapping in plants in response to abiotic stress. At present, many QTLs related to salt stress have been identified. We compared the previously reported QTLs and found that a total of 12 previously reported QTLs for ST in this study were overlapped or close to the range of previously mapped QTLs for ST.

For example, under stress conditions 151 trait–marker associations were identified that were scattered in 29 genomic regions on 10 chromosomes of rice (Nayyeripasand et al., 2021). Some of the QTLs are consistent with or close to our mapped QTL regions. The QTL qSNC1.1 on chromosome 1 coincides with the reported QTL qRDWn1.1 (root dry weight); the QTL qRNC1 is located in the region of the reported QTL qRDWs1.1 (root dry weight). The QTLs qRTRL12.1 and qRTRL12.2 on chromosome 12 are located in the region of the reported QTL qSLn12.1 (shoot length). The QTL qRTRL12.4 is 320kbp away from the previously reported QTL qRDWs12.1 (root dry weight), and the QTL qRNC12.1 (root fresh weight) is 140 kb apart from the previously reported QTL qRFWs12. The QTL qRN/K8 on chromosome 8 is located in the previously reported QTL qRFWn8.1 (root fresh weight) interval. Chen et al. (2020) identified 23 QTLs with different salt tolerance indexes through GWAS. Some of the QTLs coincide with our regions. The QTL qRSN/K1.2 is located in the region of previously reported QTL qSIS1 (salt injury score) and qSNaC1.2 (shoot Na⁺ concentration), while the QTL qSNC2 on chromosome 2 is within the interval of previously reported qRRDW2 (relative root dry weight). The QTL qSKC8 is located in the interval of previously reported qRSKC8 (relative shoot K⁺ concentration) and qRNaC8 (root Na⁺ concentration). Zhang et al. (2020) used the 55 K rice SNP array to genotype the entire population and four parents, and a total of 7 salt-tolerant QTLs were detected. Some of the QTLs coincide with or close to our regions. The QTL qRSN/K1.5 is about 174kbp away from the previously reported QTL qSLST1 (shoot length under salt stress treatment) and is located in the intervals of previously reported QTL qRDSW1 (relative dry shoot weight) and qRB1 (relative bio-mass). The QTL qSNC9 on chromosome 9 is 430 kb away from the previously reported QTL qRL-R9.1 (root length), and the QTL qRNC9 is 430kbp away from the previously reported QTL qDSW9 (dry shoot weight) and qBST9 (biomass under salt treatment). It can be seen that some of the QTLs we mapped by GWAS were overlapped with or were close to the previously reported QTLs, which demonstrates that our GWAS mapping results are rather accurate.

In addition, 52 newly identified QTLs for ST were also found in this study. Among them, qRN/K1.1 and qRSN/K1.1 were located on chromosome 1 between 12.98–13.34Mbp and can explain 13.75–16.92% of phenotypic variation. The QTL qRTRV.9 and qRTRSA.9 between 15.43–15.53Mbp on chromosome 9 can explain 13.47–17.14% of the phenotypic variation. The QTL qRN/K.9 and qRSN/K9.3 between 19.50 and 19.54 Mb on chromosome 9 can explain 16.65–21.14% of phenotypic variation. These QTLs are newly discovered and need to be further examined in the future.

Moreover, we also compared the cold-tolerant and aluminum-tolerant QTLs previously mapped using the same core collection. Some QTLs are overlapped or are similar to the QTL regions for ST in this study. Compared with the previously located QTLs for cold tolerance, the QTL qRSN/K4 in this study on chromosome 4 is about 200kbp away from the previously located cold-tolerant SNP (Chr4_20440388). The QTL qRTRV6.1 on chromosome 6 contains the previously located cold-tolerant SNP (Chr6_1320300). The QTL qRNC8.1 on chromosome 8 is about 151kbp away from the previously located cold-tolerant SNP (Chr8_374729). The QTL qRN/K10 on chromosome 10 is about 18kbp away from the previously located cold-tolerant SNP (Chr10_3365663; Song et al., 2018). Compared with the previously located QTLs for aluminum tolerance, the QTL qRN/K1.1 and qRSN/K1.1 on chromosome 1 are about 330kbp away from the previously located QTL qALT1.3. The QTL qRNC1 and qRSN/K1.3 on chromosome 1 are about 400kbp away from the previously located QTL qALT1.5. The QTL qRTRSA7 on chromosome 7 is about 84kbp away from the previously located QTL qALT7.2. The QTL region for qRSN/K9.2 contains the previously located QTL qALT9.1 (Zhao et al., 2018). These results indicate that the candidate genes in these intervals may have pleiotropic effects. It also indicates that the core collection contains a wealth of excellent resistance genes to biotic and abiotic stress.

We also compared the QTL locations with the genes known to be related to ST.⁴ Twenty-three ST genes were found to co-localize with our QTLs (Table 3). Three ST genes were found close to the mapped QTLs related to RNC in this study. Seven ST genes were found close to the mapped QTLs related to SNC. One ST gene was found in the QTL interval related to RN/K. Three ST genes were found close to the QTLs related to RTRV. Four ST genes were found close to the mapped QTLs related to RSN/K. Three ST genes were found close to the mapped QTLs related to SKC. One ST gene was found close to the QTLs related to RTRL. One ST gene was found close to the QTLs related to RTRAS. In short, these ST genes are located in or close to the relevant QTLs interval (with a searching range of +/− 250 kb). Among them, the QTL qTL4 region is composed of qSKC.6, qRN/K.6, and qSNC.6 and the overlapping regions are dense. A known ST gene OsSIDP366 was found in the region of qTL4, where a candidate gene and OsSIDP366 both contain the DUF domain. These findings support the reliability of the mapping results in this study.

The salinity tolerance in rice seedling is majorly governed by root and shoots Na⁺/K⁺ ratio. The lower Na⁺/K⁺ ratio provides protection against the toxic effects of Na⁺, hence, tolerant to salt stress. Therefore, maintaining intracellular Na⁺/K⁺ homeostasis is a key factor in determining the survival ability of plants in response to salt stress (Yang and Yan, 2018). In this study, we found that qSKC6, qSNC6, and qRN/K.6 were co-localization and 4 candidate genes (LOC_Os06g47720, LOC_Os06g47820, LOC_Os06g47850, and LOC_Os06g47970) were detected in the interval of qTL4 which are all related to ratio of Na⁺ to K⁺ concentrations; therefore, we chose the four candidate genes for further study. Besides, qRN/K1.1 and qRSN/K1.1 also were co-localization and detected in the interval of qTL6 which are related to Na⁺ and K⁺, the qTL6 contains the highest peak SNP (Chr9_19518843). Therefore, searching for candidate genes in the interval of qTL6 is also the focus of our next study.

To do it, we first checked the expression profiles of these four candidate genes from the Encyclopedia of Rice Transcriptome (TENOR) database.⁵ According to the TENOR database (Supplementary Figure S2), the candidate gene LOC_Os06g47720 has a higher expression level under salt stress conditions. The gene annotation of LOC_Os06g47720 is a serine threonine protein kinase BRI1-like 2 precursor. BRI1 (protein brassinosteroid insensitive 1) is the receptor kinase of BR (brassinosteroids), located on the cell membrane, and is a leucine-rich repeat (LRR) receptor-like serine/threonine kinase on the cell surface, which is crucial in the BR signaling pathway. BR is a plant steroid hormone, which plays a key role in growth and response to abiotic and biotic stress (Zhao et al., 2019; Ma et al., 2021). The plants adapt to various environmental stresses by changing their physiological and molecular processes, which are coordinated with changes of the levels on hormones (including brassinosteroids) in plant externally and internally (Bilal et al., 2021). The response of BR under salt stress may be mediated by BRI1, inhibiting the degradation of the endoplasmic reticulum to combat salt stress (Cui et al., 2012). BR signaling also counters salt stress by signaling cascades or initiating ethylene biosynthesis (Planas-Riverola et al., 2019). In this study, the expression level of LOC_Os06g47720 in S87 is much higher than that in S125, and LOC_Os06g47720 had far more mutation sites in S87 than in S125. By comparison of amino acid sequences, it also revealed that only the amino acids in S87 were changed and the translation was terminated in S125. These findings suggest that LOC_Os06g47720 may be involved in the regulation of ST through the BR pathway.

The candidate gene LOC_Os06g47820 is receptor-like kinases, which has the highest expression level under ABA treatment conditions (Supplementary Figure S2). ABA can coordinate with hormones such as auxin, gibberellin (GA) and cytokinin (CK) to regulate the response of plants to salt stress (Yu et al., 2020). Receptor-like kinases (RLKs) are a large family of proteins that exist on the surface of plant cell membranes. Their basic function is to transmit regulatory signals on the cell surface. Plant receptor-like protein kinases occupy important metabolic positions, and rice has about 1,130 RLK genes (Quynh-Nga et al., 2015). Plant RLKs are composed of intracellular, extracellular, and transmembrane regions (Ye et al., 2017). Receptor-like protein kinase RLK is widely involved in cell signal transduction and plant response to stress (Lemmon and Schlessinger, 2010). In recent years, several researchers studied the important role of RLKs in optimizing the response of plants to salt stress and other abiotic stresses (Zhou et al., 2018; Lin et al., 2020). In this study, the expression level of LOC_Os06g47820 in S125 was found much higher than that in S87. The gene had much more mutation sites in S125 than in S87. There are more changes in amino acids in S125 than in S87. These findings indicate that LOC_Os06g47820 is a candidate gene that may be involved in the regulation of ST in rice.

The candidate gene LOC_Os06g47850 encodes a zinc finger family protein. The zinc finger protein (ZFP) family is widely distributed in eukaryotic genomes and is one of the most important transcription factors, which plays an important role in plant growth and development and abiotic stress response (Mukhopadhyay et al., 2004; Sakamoto, 2004). More than 60 transcription factor families have been reported in plants (Iuchi, 2001). LOC_Os06g47850 has the highest expression level under cold stress conditions (Supplementary Figure S2). In our study, the expression level of LOC_Os06g47820 in S125 was much higher than that in S87. DNA sequence alignment results showed that the gene has no mutation sites. Therefore, the mechanism of this candidate gene remains to be elucidated.

The candidate gene LOC_Os06g47970 encodes a DUF1517 (domains of unknown function, DUF) which are a class of proteins whose functions have not been characterized and account for about 25% of the total protein family (Mudgal et al., 2015). According to the TENOR database (Supplementary Figure S2), LOC_Os06g47970 has the highest expression level under drought stress and as we know that salinity usually occurs at the same time as drought stress (Hu et al., 2006). In recent years, an increasing number of studies were conducted on the regulation of different DUFs family genes involved in plant growth and development and plant response to stress (biotic and abiotic stress; Li et al., 2018; Lv et al., 2019). We found a DUFs gene in the qTL4, i.e., OsSIDP366, which is a stress-induced DUF1644 protein and contains a DUF1644 domain, a C2H2 and a ring finger domain. OsSIDP366 is expressed in multiple tissues. The expression is higher in young roots, mature leaves, and leaf sheaths, but lower or no expression in internodes, mature seeds, lemmas and glumes. High salt and drought treatments can induce the expression of OsSIDP366. OsSIDP366 may positively regulate salt and drought resistance in rice (Guo et al., 2016). In addition, we also found that the homologous gene AT5G57345 in Arabidopsis, a single copy gene, is localized to ER and expressed in the whole plant and induced expression in response to abiotic stress. Although the function of AT5G57345 is unclear, overexpression can lead to increased tolerance to abiotic stress and increased ascorbic acid content (Bu et al., 2016). In addition, we also found that the expression level of LOC_Os06g47970 in S125 was much higher than that in S87. Sequence analysis found that LOC_Os06g47970 only had a pair of base substitutions in S125, i.e., one amino acid was changed in S125, while no change in S87. Therefore, LOC_Os06g47970 might be involved in the regulation of ST in rice.

This study lays a foundation for the functional analysis of the candidate genes in the regulation of salt tolerance in rice and the enrichment of the rice salt tolerance regulatory network. In addition, our newly discovered QTLs also lay the foundation for further research on the ST mechanism in rice in the future. The salt tolerance-related candidate genes and QTLs would provide important resources for molecular breeding and functional analysis of the salt tolerance during the seedling stage of rice. However, the four candidate genes identified in this study which involved in the regulation of salt stress in rice need further research and verification. In addition, the mechanism and the regulation pathway for these genes under salt stress in rice still need to be clarified. In future, the function of candidate genes related to salt tolerance will be studied by Crispr-Cas9 technology, which will help to precisely uncover the mechanisms of salinity tolerance at molecular level.